What is the method for calculating residuals for 2 variable data?

[pyroknife]

Let's say I have a 2 variable data set x and y.
The linear fit through the line is some y(x). Let's say y=x for simplicity.

So a residual is defined as the actual value minus the estimated value (using y).

My question is, let's say for one x value there are two y values.
e.g., for x=2; y=4 and y=6

If you wanted to calculate the residual at x=2, would you have two separate values or is there someway to combine it?
The estimated value would be y=2.
residuals are: 2-4 and 2-6, which is -2 and -4. So in other words do you add the two values?

 

[haruspex]

You should count them as two separate residuals. To do otherwise would overestimate the goodness of fit.